A

Campcw©

K

1

ettematrozd

e

Hf

(

0d

Journal of Membrane Science 297 (2007) 236–242

Novel ploy(vinyl alcohol)/carbon nanotube hybrid membranes forpervaporation separation of benzene/cyclohexane mixtures

Fubing Peng a,b,∗, Changlai Hu b, Zhongyi Jiang b,∗∗a College of Chemistry and Chemical Engineering, Hunan Normal University, Changsha 410081, PR China

b Key Laboratory for Green Chemical Technology, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, PR China

Received 2 February 2007; received in revised form 24 March 2007; accepted 27 March 2007Available online 31 March 2007

bstract

Novel ploy(vinyl alcohol)/carbon nanotube hybrid membranes were prepared and carbon nanotube was dispersed by using �-cyclodextrin (�-D). These hybrid membranes were characterized by TEM, SEM and DMA. Both pure PVA and �-CD-CNT/PVA hybrid membranes are uniformnd these hybrid membranes exhibited significant improvement in Young’s modulus and thermal stability as compared to pure PVA and �-CD/PVAembranes. These membranes were applied to pervaporation separation of benzene/cyclohexane mixtures, and showed excellent pervaporation

roperties. The permeation flux of benzene could be 61.0 g/(m2 h) and separation factor could be 41.2, which are above the upper bound trade-offurve summarized by Lue and Peng. The effects of �-CD-CNT content, operating temperature and feed flow rate on pervaporation properties alsoere investigated.2007 Elsevier B.V. All rights reserved.

brane

mpmiomtpnrm[f

eywords: Pervaporation; Poly(vinyl alcohol); Carbon nanotube; Hybrid mem

. Introduction

Organic mixtures have conventionally been separated byxtractive distillation, extraction and adsorption processes; buthere are high capital investment and energy consumption forhese separation technologies. Recently, pervaporation, as annvironment-benign and energy-saving technology, has gaineduch attention to separate organic mixtures due to its high sep-

ration efficiencies coupled with energy savings, especially forhe close boiling point and azeotropic mixtures [1,2]. Pervapo-ation is a very promising membrane technology for separationf organic/organic mixtures, among which separating ben-ene/cyclohexane mixture is one of the most important and most

ifficult processes.

Organic–inorganic hybrid membranes have attracted consid-rable recent attention as potential “next generation” membrane

∗ Corresponding author at: College of Chemistry and Chemical Engineering,unan Normal University, Changsha 410081, PR China. Tel.: +86 731 8872238;

ax: +86 731 8872238.∗∗ Corresponding author. Tel.: +86 22 27892143; fax: +86 22 27892143.

E-mail addresses: fbpeng@gmail.com (F. Peng), zhyjiang@tju.edu.cnZ. Jiang).

b

amoblbba

376-7388/$ – see front matter © 2007 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2007.03.048

; �-Cyclodextrin

aterials [3–5]. Such hybrid membranes are typically com-osed of polymeric and inorganic materials, and have bothembrane-forming properties of a polymer and physicochem-

cal stability of an inorganics [6,7]. Inorganic particles inrganic–inorganic hybrid membrane are usually silica, zeolite,etal oxide nanoparticle, nanotube and so on. More recently,

here has been growing interest in exploring new applications oforous carbons because of their ability to interact with moleculesot only at their surfaces but also within the bulk of the mate-ial [8]. In our previous paper, we had reported PDMS-carbonolecular sieve (CMS) [9], poly(vinyl alcohol) (PVA)–CMS

10], and PVA–graphite composite or hybrid membranes [2,11]or removal of benzene from aqueous solution and separation ofenzene/cyclohexane mixtures.

Poly(vinyl alcohol) (PVA) is polar and hydrophilic, and isn ideal membrane material to separate benzene/cyclohexaneixture [12] due to the distinct preferential adsorption/solution

f PVA toward benzene over cyclohexane since the solubility ofenzene in water (1.8 g/L, 298 K [13]) is one order of magnitude

arger than that of cyclohexane (0.0561 g/L, 298 K [20]). Butecause of the big difference of solubility parameters betweenenzene/cyclohexane and PVA, PVA often showed lower perme-bility to benzene and cyclohexane. Considering the superior

brane

pealch

otbrfec

hdtcCp

2

2

1jcLfSuu

2

cigtth

accac4wbaoa

o0

2

nJapebwPtam

2

vmsirsccgpvpctsu

J

α

HiPtws

3

3

F. Peng et al. / Journal of Mem

roperties such as high flexibility, low mass density, plus theffective �–� stacking interaction between carbon nanotube andromatic compounds [14,15], carbon nanotube (CNT) is specu-ated to be excellent candidate for substituting or complementingonventional nanofillers in the fabrication of organic–inorganicybrid pervaporation membrane [16–19].

To fully explore the potential of CNT, the serious aggregationf CNTs leads to difficulties in their manipulation and incorpora-ion into polymeric matrixes [20]. The dispersion and solubilityehavior of CNT can be remarkably improved through incorpo-ation of cyclodextrin [21,22]. In our previous study, we haveound that incorporation of �-CD into PVA matrix could consid-rably enhance the pervaporation performance of benzene andyclohexane mixtures [23].

In this study, the pervaporation properties of poly(vinyl alco-ol) (PVA) membranes filled with carbon nanotubes (CNT)ispersed by �-CD for separation of benzene/cyclohexane mix-ures were studied. The membrane structure and properties wereharacterized by TEM, SEM and DMA. The effects of �-CD-NT content, operating temperature and feed flow rate on theervaporation properties of these membranes were investigated.

. Experimental

.1. Materials

Poly(vinyl alcohol) (the degree of polymerization was750 ± 50, degree of hydrolysis was 95%) was supplied by Tian-in Yuanli Chemical Company (Tianjin, China), benzene andyclohexane were purchased from Tianjin Jiangtian Chemicalstd., Tianjin, China. Multi-wall carbon nanotube was purchased

rom Tsinghua University, China. �-CD was purchased fromigma Co. All the chemicals were of analytical grade and weresed without further purification. Double distilled water wassed throughout the study.

.2. Membrane preparation

The CNT was dispersed by �-CD, and the preparation pro-edure was in accordance with that outlined by Chen et al. [22]n which a mixture of �-CD and CNT (in a 30:1 ratio) wasround in an agate mortar and pestle for approximately 2 h withhe dropwise addition of ethanol (1 mL) over the first 10 min,hen dried for 24 h at 348 K. This procedure resulted in a fineomogeneous black powder (�-CD-CNT).

Poly(vinyl alcohol) (5 g) was dissolved in 45 g distilled watert 363 K. The hot solution was filtered and to the filtrate, aertain amount carbon nanotube dispersed by �-CD and 2 mLross-linker glutaraldehyde (25 wt.% aqueous solution) as wells 1 mL of concentrated HCl catalyst were added to initiate theross-linking reaction. The solution was gently stirred for abouth at room temperature and the resulting homogeneous solutionas cast onto a glass plate with the casting knife. The mem-

ranes were allowed to dry at room temperature for 1–2 daysnd the completely dried membranes were subsequently peeledff. Then, the membranes were heat-treated at 393 K for 1–3 h,nd the membrane thickness was about 80 �m. The mass ratio

3

d

Science 297 (2007) 236–242 237

f �-CD-CNT to PVA was varied as 0.02, 0.04, 0.06, 0.08, 0.10,.20 and 0.30.

.3. Membrane characterization

The structures of the pristiline carbon nanotubes and carbonanotubes dispersed by �-CD were examined by a JEOL-EM-100CX � transmission electron microscopy (TEM) withn acceleration voltage of 1.0 × 105 V. The cross-section mor-hologies of the membranes were investigated by scanninglectron microscopy (SEM) (XL30ESEM, PHILIPS). The mem-rane samples were fractured in liquid nitrogen and then coatedith gold. Dynamic mechanical data were obtained with aerkin-Elmer DMA instrument. All samples were tested within

he temperature range of 298–500 K at a heating rate of 5 K/min,nd a frequency of 1 Hz was selected for all the experi-ents.

.4. Pervaporation experiment

Pervaporation experiment apparatus is shown in Fig. 1. Per-aporation experiments were performed on the P-28 membraneodule (CM-Celfa AG Company, Switzerland). The effective

urface area of the membrane in contact with the feed mixtures 28.0 cm2. The vacuum in the downstream side of the appa-atus was maintained (1 kPa) using a vacuum pump. After ateady state (about 2 h) was attained, the permeate liquid wasollected in cold traps immersed in the liquid nitrogen. Theompositions of benzene and cyclohexane were estimated byas chromatography (Agilent 6820, USA). The results from theermeation of benzene/cyclohexane mixtures during the per-aporation were reproducible, and the errors inherent in theervaporation measurements were in the order of a few per-ent. From the pervaporation data, separation performances ofhe membranes can be assessed in terms of total flux (J) andeparation factor (αPV), and they were calculated, respectively,sing the following equations:

= W

At(1)

PV = PB/PC

FB/FC(2)

ere W is the mass of permeate (g), A the area of the membranen contact with the feed mixture (m2), t the permeation time (h),B and PC the weight fractions of benzene and cyclohexane in

he permeate, respectively, and FB and FC are the respectiveeight fractions of benzene and cyclohexane in the feed. In this

tudy, benzene/cyclohexane (50/50 wt.%) mixtures were used.

. Results and discussion

.1. Membrane characterization

.1.1. Transmission electron microscopy (TEM)Transmission electron microscopy (TEM) was used to

irectly view the structure of the pristiline carbon nanotubes and

238 F. Peng et al. / Journal of Membrane Science 297 (2007) 236–242

F 2) feec gen t

tpAg

F

s

ig. 1. The schematic diagram of the PV experimental equipment. (1) Heater; (ell; (7) membrane, (8) rotameter, (9) permeate collection tube; (10) liquid nitro

he carbon nanotubes dispersed by �-CD to investigate the dis-

ersion result. Typical TEM images were shown in Figs. 2 and 3.s shown in Fig. 3, the dispersion of CNT by �-CD was veryood. CNT and �-CD do interact with each other in a fashion

Fig. 2. The TEM photographs of the pristine carbon nanotubes.

ig. 3. The TEM photographs of the carbon nanotubes dispersed by �-CD.

pb

3

trhCmPbwwhlP

3

Pb5mptTtmatPbphrp

d tank; (3) liquid level meter; (4) thermocouple; (5) feed pump; (6) membranerap; (11) vacuum pump.

imilar to the originally suggested by Chen et al. [22], who pro-osed that �-CD was absorbed at the surface of nanotube ropesy van der Waals forces.

.1.2. Scanning electron microscopy (SEM)In order to investigate the membrane structure, SEM charac-

erization of these membranes has been carried out. The SEMesults of the cross-section of pure PVA and �-CD-CNT/PVAybrid membranes are shown in Fig. 4. Pure PVA and �-CD-NT/PVA hybrid membranes were dense with no connectedacroscopic voids and CNT dispersed uniformly within theVA matrix. The aggregation of CNT was obviously improvedy incorporation of �-CD compared with pristine CNT, whichas difficult to make a uniform membrane. The �-CD-CNT wasell adhesive with PVA due to the good compatibility betweenydrophilic �-CD and hydrophilic PVA. Therefore, no nonse-ective defects voids could be found at the interface between theVA and CNT additives.

.1.3. Dynamic mechanical analysis (DMA)Fig. 5 shows the dynamic mechanical behavior of pure

VA membrane, �-CD/PVA and �-CD-CNT/PVA hybrid mem-ranes. The tan δ curves in the temperature range from 300 to00 K of pure PVA, �-CD/PVA, and �-CD-CNT/PVA hybridembranes are shown in Fig. 5(a). There are two peaks for

ure PVA and �-CD/PVA membrane, and the relaxation transi-ions are usually labeled by � and � with deceasing temperature.he highest transition, the � relaxation, is the glass-to-rubber

ransition (Tg). Tg was decreased from 375 K for pure PVAembrane to 340 K for �-CD-CNT/PVA hybrid membrane,

nd Tg for �-CD/PVA membrane was 468 K. Fig. 5(b) showshe Young’s modulus as a function of the temperature for pureVA membrane, �-CD/PVA and �-CD-CNT/PVA hybrid mem-

ranes. Among all the test membranes, the Young’s modulus ofure PVA membrane was the lowest and that of �-CD-CNT/PVAybrid membrane was the largest, which indicated that incorpo-ating with �-CD and CNT both strengthened the mechanicalroperties of pure PVA membrane.

F. Peng et al. / Journal of Membrane Science 297 (2007) 236–242 239

l) me

3s

mb

FC

Cw

Fig. 4. The SEM photographs of the poly(vinyl alcoho

.2. Effect of β-CD/CNT content on permeation flux andeparation factor

Fig. 6 shows the effect of �-CD-CNT content on the per-eation flux and separation factor. The permeation flux of

enzene and total permeation flux were increased when the �-

ig. 5. The dynamic mechanical properties results of PVA, �-CD/PVA and �-D-CNT/PVA (8% �-CD-CNT) membranes.

tmsotoprtCsoowhatiWor

3s

tctfpr3s

3s

zr

mbrane: (a) pure PVA and (b) 8% �-CD-CNT (30/1).

D-CNT content increased from 0 to 6 wt.%, and decreasedhen �-CD-CNT content increased from 6 to 30 wt.%. Both

otal permeate flux and permeation flux of benzene reached theaximal value when �-CD-CNT content was 6 wt.%. The rea-

ons were, firstly, �-CD-CNT could enhance sorption selectivityf PVA membrane to benzene compared with cyclohexane dueo the �–� interaction between CNT and benzene molecules;n the other hand, the distribution of �-CD-CNT to interfereolymer chain packing in the membrane increased the transportesistance of benzene and cyclohexane molecules. The separa-ion factor increased with the �-CD-CNT content increasing.omparing with these three different membranes, it could be

een that the permeation flux of benzene and separation factorf �-CD-CNT/PVA hybrid membranes were higher than thatf pure PVA membranes and �-CD-PVA membranes, whichas due to the special interaction between CNT and aromaticydrocarbon compounds. There are a lot of � bonds in the CNTnd they have interaction with the � bonds that benzene con-ains. This interaction could enhance the adsorption of benzenen the membrane surface and the diffusion in the membrane.

hen the �-CD-CNT content is 6 wt.%, the permeation fluxf benzene and separation factor are 42.3 g/(m2 h) and 36.4,espectively.

.3. Effect of operating temperature on permeation flux andeparation factor

The effect of operating temperature on pervaporation proper-ies of �-CD-CNT/PVA hybrid membranes is shown in Fig. 7. Itould be observed that both permeation flux and separation fac-or increased significantly with operating temperature increasingrom 303 to 333 K. It could be also seen that the relationship ofermeation flux and separation factor obeyed the Arrhenius-typeelationship from Fig. 7(a). When the operating temperature was33 K, permeation flux of benzene could be 60.0 g/(m2 h) andeparation factor could be 41.2.

.4. Effect of feed flow rate on permeation flux andeparation factor

As shown in Fig. 8, both total permeation flux and ben-ene permeation flux increased with the increase of feed flowate, whereas permeation flux of cyclohexane changed slightly.

240 F. Peng et al. / Journal of Membrane Science 297 (2007) 236–242

Fig. 6. The effect of �-CD-CNT content on pervaporation properties of �-CD-CNT/PVA hybrid membrane (feed temperature: 323 K, feed concentration: 50 wt.%benzene, feed flow rate: 50 mL/min).

Fig. 7. The effect of feed temperature on pervaporation properties of �-CD-CNT/PVA hybrid membrane (�-CD-CNT content: 8 wt.%, feed concentration: 50 wt.%benzene, feed flow rate: 50 mL/min).

Fig. 8. The effect of feed flow rate on the pervaporation properties of �-CD-CNT/PVA hybrid membrane (feed temperature: 323 K, feed concentration: 50 wt.%benzene, �-CD-CNT content: 8 wt.%).

brane

Mfl

iemblSflfliob

3m

etsoiomtbuuror�t

Fi(

4

�thozuTmt

A

CPiX

R

F. Peng et al. / Journal of Mem

eanwhile, separation factor increased with the increase of feedow rate.

With feed flow rate increasing, the turbulence of feed solutionncreased and the thickness of boundary layer decreased, i.e., thextent of concentration polarization on the upstream surface ofembranes decreased. Therefore, mass transfer resistance of

oundary layer on the upstream of membrane decreased, whiched to the increase of total permeation flux correspondingly.ince permeation flux of benzene increased while permeationux of cyclohexane decreased a little with increasing of feedow rate, separation factor increased with feed flow rate increas-

ng. When feed flow rate was 110 mL/min, the permeation fluxf benzene could be 52.4 g/(m2 h) and separation factor coulde 50.2.

.5. Comparison of pervaporation properties of differentembranes for separating benzene/cyclohexane mixtures

The intrinsic trade-off between permeability and selectivityxists in polymeric membranes. Robeson [12] has summarizedhe “upper bound trade-off curve” for O2/N2, CO2/CH4, etc. gaseparation polymeric membranes, which initiate the researchf polymeric membrane with high permeability and selectiv-ty. Lue and Peng [24] have collected the pervaporation resultsn benzene/cyclohexane separation using various membraneaterials on 50–60 wt.% benzene composition at an operating

emperature of 20–80 ◦C and induced the relationship betweenenzene/cyclohexane selectivity and benzene flux to plot anpper bound trade-off curve shown in Fig. 9. According to thispper bound trade-off curve, in our experiment, the pervapo-ation properties of pure PVA and �-CD/PVA membranes was

bviously below this upper bound trade-off curve. By incorpo-ating CNT into PVA membrane, pervaporation properties of-CD-CNT/PVA hybrid membranes are above the upper bound

rade-off curve.

ig. 9. Relationship between the benzene/cyclohexane selectivity and normal-zed permeation flux of different membranes. Normalized permeation fluxkg �m/(m2 h)) = permeation flux (kg/(m2 h)) × membrane thickness (80 �m).

[

[

[

[[

[

Science 297 (2007) 236–242 241

. Conclusions

Incorporation of carbon nanotube, which was dispersed with-CD by grinding method, into PVA membranes could increase

he permeate flux and separation factor. The �-CD-CNT/PVAybrid membrane exhibited the highest benzene permeation fluxf 61.0 g/(m2 h) with separation factor of 41.2 at 333 K for ben-ene/cyclohexane (50/50 wt.%) mixtures, which are above thepper bound trade-off curve summarized by Lue and Peng.he �-CD-CNT/PVA membranes demonstrated much betterechanical strength properties, sorption and diffusion properties

han pure PVA and �-CD/PVA membranes.

cknowledgments

We gratefully acknowledge financial support from the Cross-entury Talent Raising Program (Ministry of Education, China),rogram for Changjiang Scholars and Innovative Research Team

n University (PCSIRT), and SINOPEC Research Program (NO.503029).

eferences

[1] J. Neel, in: R.Y.M. Huang (Ed.), Pervaporation Membrane Separation Pro-cess, 1991 (Chapter 1).

[2] F.B. Peng, L.Y. Lu, C.L. Hu, H. Wu, Z.Y. Jiang, Significant increaseof permeation flux and selectivity of poly(vinyl alcohol) membranes byincorporation of crystalline flake graphite, J. Membr. Sci. 259 (2005)65.

[3] T.C. Merkel, Z. He, I. Pinnau, Effect of nanoparticles on gas sorption andtransport in poly(1-trimethylsilyl-1-propyne), Macromolecules 36 (2003)6844.

[4] T. Uragami, K. Okazaki, H. Matsugi, T. Miyata, Structure and perme-ation characteristics of an aqueous ethanol solution of organic–inorganichybrid membranes composed of poly(vinyl alcohol) and tetraethoxysilane,Macromolecules 35 (2002) 9156.

[5] M. Smaihi, T. Jermoumi, J. Marignan, R.D. Noble, Organic–inorganic gasseparation membranes: preparation and characterization, J. Membr. Sci.116 (1996) 211.

[6] K.A. Mauritz, J.T. Payne, [Perfluorosulfonate ionomer]/silicate hybridmembranes via base-catalyzed in situ sol–gel processes for tetraethy-lorthosilicate, J. Membr. Sci. 168 (2000) 39.

[7] C. Hibshman, C.J. Cornelius, E. Marand, The gas separation effects ofannealing polyimide–organosilicate hybrid membranes, J. Membr. Sci. 211(2003) 25.

[8] B. Sakintuna, Y. Yurum, Templated Porous Carbons: A Review Article,Ind. Eng. Chem. Res. 44 (2005) 2893.

[9] F.B. Peng, Z.Y. Jiang, C.L. Hu, Y.Q. Wang, H.Q. Xu, J.Q. Liu, Removingbenzene from aqueous solution using CMS-filled PDMS pervaporationmembranes, Sep. Purif. Technol. 48 (2006) 229.

10] H.L. Sun, L.Y. Lu, F.B. Peng, H. Wu, Z.Y. Jiang, Pervaporation ofbenzene/cyclohexane mixtures through CMS-filled poly(vinyl alcohol)membranes, Sep. Purif. Technol. 52 (2006) 203.

11] F.B. Peng, L.Y. Lu, H.L. Sun, Z.Y. Jiang, Organic–inorganic hybrid mem-branes with simultaneously enhanced flux and selectivity, Ind. Eng. Chem.Res. 46 (2007) 2544.

12] L.M. Robeson, Correlation of separation factor versus permeability forpolymeric membranes, J. Membr. Sci. 62 (1991) 165.

13] C.L. Yaws, Chemical Properties Handbook, McGraw-Hill Book Co., 1999.

14] C.A. Hunter, J.K.M. Sanders, The nature of �–� interactions, J. Am. Chem.

Soc. 112 (1990) 5525.15] J. Zhao, J.P. Lu, J. Han, C.-K. Yang, Noncovalent functionalization of

carbon nanotubes by organic molecules, Appl. Phys. Lett. 82 (2003)3746.

2 brane

[

[

[

[

[

[

[

[

membranes without and with �-cyclodextrin, Desalination 193 (2006)

42 F. Peng et al. / Journal of Mem

16] J.H. Sung, H.S. Kim, H.-J. Jin, H.J. Choi, I.-J. Chin, Nanofi-brous membranes prepared by multiwalled carbon nanotube/poly(methylmethacrylate) composites, Macromolecules 37 (2004) 9899.

17] Y. Lin, B. Zhou, K.A. Fernando, P. Liu, L.F. Allard, Y.-P. Sun, Polymericcarbon nanocomposites from carbon nanotubes functionalized with matrixpolymer, Macromolecules 36 (2003) 7199.

18] N. Grossiord, J. Loos, O. Regev, C.E. Koning, Toolbox for dispersing car-bon nanotubes into polymers to get conductive nanocomposites, Chem.Mater. 18 (2006) 1089.

19] M. Moniruzzaman, K.I. Winey, Polymer nanocomposites containing car-bon nanotubes, Macromolecules 39 (2006) 5194.

20] D.M. Delozier, K.A. Watson, J.G. Smith, T.C. Clancy, J.W. Connell, Opti-cally active photochromic polymers with three-arm star structure by atomtransfer radical polymerization, Macromolecules 39 (2006) 1731.

[

Science 297 (2007) 236–242

21] G. Chamber, C. Carroll, G.F. Farrell, Characterization of the interactionof gamma cyclodextrin with single-walled carbon nanotubes, Nano Lett. 3(2003) 843.

22] J. Chen, M.J. Dyer, M.-F. Yu, Cyclodextrin-mediated soft cuttingof single-walled carbon nanotubes, J. Am. Chem. Soc. 123 (2001)6201.

23] F.B. Peng, Z.Y. Jiang, C.L. Hu, Y.Q. Wang, L.Y. Lu, H. Wu, Perva-poration of benzene/cyclohexane mixtures through poly(vinyl alcohol)

182.24] S.J. Lue, S.H. Peng, Polyurethane (PU) membrane preparation with and

without hydroxypropyl-�-cyclodextrin and their pervaporation character-istics, J. Membr. Sci. 222 (2003) 203.